Semiconductor die have traditionally been electrically connected to a package by wire bonding techniques, in which wires are attached to pads of the die and to pads located in the cavity of the plastic or ceramic package. Wire bonding is still the interconnection strategy most often used in the semiconductor industry today. But the growing demand for products that are smaller, faster, less expensive, more reliable and have a reduced thermal profile has pushed wire bonding technology to its limits (and beyond) thereby creating barriers to sustained product improvement and growth.
The high-performance alternative to wire bonding techniques are flip chip techniques, in which solder balls or bumps are attached to the input/output (I/O) pads of the die at the wafer level. The bumped die is flipped over and attached to a substrate “face down,” rather than “face up” as with wire bonding. Flip chips resolve many if not all of the problems introduced by wire bonding. First, flip chips have fewer electrical interconnects than wire bonding, which results in improved reliability and fewer manufacturing steps, thereby reducing production costs. Second, the face down mounting of a flip chip die on a substrate allows superior thermal management techniques to be deployed than those available in wire bonding. Third, flip chips allow I/O to be located essentially anywhere on the die, within the limits of substrate pitch technology and manufacturing equipment, instead of forcing I/O to the peripheral of the die as in wire bonding. This results in increased I/O density and system miniaturization.
Despite the advantages of the flip chip, wide spread commercial acceptance of the flip chip has been hindered by testing issues. To ensure proper performance, the die should be adequately tested before it is assembled into a product; otherwise, manufacturing yields at the module and system level can suffer and be unacceptably low. Under some circumstances, a defective die can force an entire subassembly to be scrapped. One attempt to address this testing issue has been to perform a wafer probe, followed by dicing the wafer and temporarily packaging each die into a test fixture of some sort. Performance testing is subsequently executed. Burn-in testing is often included in this process to eliminate any die having manufacturing process defects. Following the successful completion of these tests, the die are removed from the test fixture and either retailed as a Known Good Die (“KGD”) product or used by the manufacturer in an end product, such as a Multichip Module (“MCM”). The Multichip Module may constitute a subassembly in a larger system product. This Known Good Die process is inherently inefficient due to its complexity.
Accordingly, there is a need for a system, method and apparatus for testing multiple semiconductor wafers that is simple, allows testing at the wafer level before dicing, and eliminates the need for temporarily packaging the die in a carrier.